- Email: [email protected]
Predictors of Mortality After Surgical Management of Lung Cancer in the National Cancer Database
Predictors of Mortality After Surgical Management of Lung Cancer in the National Cancer Database Joshua E. Rosen, BASc, Jacquelyn G. Hancock, BS, Anthony W. Kim, MD, Frank C. Detterbeck, MD, and Daniel J. Boffa, MD Department of Thoracic Surgery, Yale University School of Medicine, New Haven, Connecticut
Background. Surgical resection represents the standard of care for locoregionally confined non-small cell lung cancer (NSCLC); however, surgical complications may compromise the overall outcome. Adverse events after lung cancer surgery have been studied extensively, yet available databases have significant limitations (with respect to size, provider, patient age, and so forth). The National Cancer Database (NCDB) is the largest cancer registry in the world, capturing 67% of newly diagnosed NSCLC in the United States. We studied surgically managed NSCLC patients in the NCDB to more accurately assess factors that influence perioperative outcomes. Methods. Patients diagnosed with NSCLC from 2004 to 2009 in the NCDB who were managed with surgical resection were included (n [ 119,146). Primary endpoints included death within 30 days of surgery and extended length of stay (more than 14 days). Results. Overall 30-day mortality rate was 3.4% and varied by procedure: lobectomy/bilobectomy (2.6%),
wedge resection (4.2%), extended lobectomy/bilobectomy (4%), and pneumonectomy (8.5%). By multivariable analysis, increasing age, male sex, increasing comorbidities, and decreased facility volume were associated with higher 30-day mortality. Of patients who underwent lobectomy, 9.1% had an extended length of stay. On multivariable analysis, increasing age, male sex, increasing comorbidities, decreasing facility volume, and right-sided tumors were associated with increased incidence of extended length of stay. Conclusions. Adverse events after the surgical treatment of NSCLC in the NCDB occur with a similar frequency and are predicted by similar patient, procedural, and facility variables as have been identified by more restricted data resources. The NCDB appears to be a valuable resource to study NSCLC in the United States.
L
28% of the cancers diagnosed in the United States) [9, 10]. Furthermore, the NCDB population lacks many of the restrictions affecting other administrative and clinical databases (eg, SEER Medicare is limited to patients over the age of 65 years [9] and The Society of Thoracic Surgeons (STS) database is almost entirely limited to cases performed by board-certified thoracic surgeons) [11]. Therefore, the NCDB represents an important perspective on the management of NSCLC in the United States. We studied 30-day mortality and extended length of stay (eLOS), as a surrogate for significant but nonlethal complications, in the NCDB to better understand the factors that influence perioperative outcomes in surgically managed NSCLC in the United States.
ung cancer is the leading cause of cancer-related mortality in the United States, claiming more than 160,000 lives in 2010 alone [1]. Surgical resection plays an important role in the management of locoregionally confined non-small cell lung cancer (NSCLC) [2–6]; however, the benefits of surgery may be offset by adverse events in the postoperative period. As a result, short-term surgical outcomes such as death and nonlethal complications are important considerations in comparative effectiveness studies for the management of locoregionally confined NSCLC. Adverse events after lung cancer surgery have been studied extensively, but with mixed, often conflicting results [7, 8]. Because patients, tumors, and cancer providers vary considerably, it is possible that variability in short-term outcome studies to some degree reflects heterogeneity in the available data resources. The National Cancer Data Base (NCDB) represents the largest cancer database in the United States, capturing an estimated 67% of newly diagnosed lung cancers (by comparison, the Surveillance, Epidemiology and End Results [SEER] database only captures approximately Accepted for publication July 7, 2014. Address correspondence to Dr Boffa, 330 Cedar St, BB205, PO Box 208062, New Haven, CT 06520-8062; e-mail: [email protected].
Ó 2014 by The Society of Thoracic Surgeons Published by Elsevier
(Ann Thorac Surg 2014;-:-–-) Ó 2014 by The Society of Thoracic Surgeons
Patients and Methods Data Source The NCDB is a hospital-based tumor registry jointly run by the American College of Surgeons and American Cancer Society since 1985. Currently, all Commission on Cancer accredited facilities are mandated to report all newly diagnosed cancer cases to the NCDB. These approximately 1,445 facilities comprise around 30% of the hospitals in the United States. It is estimated that 67% of 0003-4975/$36.00 http://dx.doi.org/10.1016/j.athoracsur.2014.07.007
2
ROSEN ET AL LUNG CANCER SURGERY IN THE NCDB
lung and bronchus cancers in the United States are captured by the NCDB [9]. The data used in this study are derived from a deidentified NCDB file. The American College of Surgeons and the Commission on Cancer have not verified and are not responsible for the analytic or statistical methodology employed, or the conclusions drawn from these data by the investigator. This study was approved by the Institutional Review Board of the Yale School of Medicine.
Patient Selection The NCDB lung-cancer beta participant user file was queried for all patients over the age of 19 years diagnosed with NSCLC from 2004 to 2009 (n ¼ 668,264). Eligible patients included those undergoing surgical resection (wedge resection, segmentectomy, lobectomy/bilobectomy, extended lobectomy/bilobectomy, pneumonectomy [n ¼ 183,341]). Patients were excluded if they had a noninvasive cancer (n ¼ 133), history of previous cancers (n ¼ 61,370), underwent noncurative intent therapy (n ¼ 1,754), participated in an experimental protocol (n ¼ 404), received intraoperative radiation (n ¼ 728), received hormone therapy (n ¼ 4,100) or immunotherapy (n ¼ 3,118), had an unknown TNM edition number (n ¼ 210), had a pathologic stage of 0 or occult (n ¼ 396), or had missing 30-day mortality data (n ¼ 138). In all, 119,146 patient records met the selection criteria and were used for further analysis. Patients with missing outcome data for length of stay (LOS) were dropped from the analysis of that outcome; furthermore, patients with missing data for neoadjuvant treatment (n ¼ 7,106) were not included in the multivariable models after analyzing for significant confounding differences (data available on request). Additionally, we performed a subsequent analysis of 27,563 separate lobectomy/bilobectomy patients from the 2010 to 2011 NCDB cohort to determine the impact of operative approach on 30-day mortality and eLOS (approach was only added to the data file in 2010). Cases that included both minimally invasive and open approaches were excluded because it is unclear whether they represent a planned thoracoscopic procedure for diagnosis followed by open lobectomy/bilobectomy, or true conversion due to operative findings or complications.
Data Elements The dependent variables included death within 30 days of the primary surgical procedure, and eLOS after the primary surgical procedure; eLOS was defined as more than 14 days from the date of surgery [12]. Independent variables included year of diagnosis, age, sex, race, insurance type, income level, modified Charlson-Deyo score, facility location, facility type, histology, primary tumor site, tumor laterality, pathological stage group, surgical procedure, and neoadjuvant treatment use. Facility volume data for each patient was calculated as the total volume of procedures performed at the treating facility in the year of the patient’s diagnosis, to account for variations in facility volume over time. Comorbidities were represented as a modified Charlson-Deyo score (only includes 0, 1, and 2). Missing data for each variable were coded as
Ann Thorac Surg 2014;-:-–-
“unknown” for the multivariable models, as has been done in similar studies [8]. The use of neoadjuvant radiotherapy was determined by using a specific data field in the NCDB that specifies the temporal relationship of the radiotherapy treatment to the primary surgical procedure. A similar code for systemic therapy was not put in place until 2006. The use of neoadjuvant chemotherapy was determined by comparing the number of days elapsed from diagnosis to surgery, and the number of days elapsed from diagnosis to the start of chemotherapy.
Statistical Analysis Bivariate analysis was performed using the c2 test for categorical covariates and one-way analysis of variance for continuous covariates. Of the 119,146 patients included, 10,887 (9%) were missing data on length of stay and were not included in models of eLOS. Multivariable risk models were developed for each endpoint (30-day mortality, incidence of eLOS). Model development was a multistep process, beginning with simple logistic regression using covariates chosen based on clinical interest and significance in the bivariate analysis (p value threshold < 0.05). Variables were inspected for colinearity. We then transitioned to using generalized linear mixed models using a binomial distribution and logit link with random effects to account for patient clustering within hospitals to calculate parameter estimates and p values [13]. All data elements, except for age and facility volume, were treated as nominal covariates, whereas age and facility volume were analyzed as continuous covariates. Model data are reported using odds ratio (OR), 95% confidence interval (CI), and p values. Model discrimination was assessed using the c-statistic, which was 0.75 for the 30-day mortality model and 0.68 for the eLOS model. All analysis was performed using SAS, version 9.3 (SAS Institute, Cary, NC). A p value of less than 0.05 was set as the threshold for significance.
Results Patient Population In all, 119,146 cases of surgically managed lung cancer were identified and the demographic, cancer, and operative characteristics are described in Table 1. The median age at diagnosis was 67 years, with an interquartile range of 60 to 74. Patients were generally healthy (87% had one or no comorbidities). Most patients underwent lobectomy/bilobectomy (76%), followed by wedge resection (14%), pneumonectomy (7%), and segmentectomy (3%). The median procedure volume of the treating facility was 11 cases per year, with an interquartile range of 5 to 21 cases per year.
Thirty-Day Mortality Overall, 3.4% of patients died within 30 days of their lung cancer surgery, decreasing over the study period from 3.8% in 2004 to 3.0% in 2009 (p < 0.0001). Unadjusted
Ann Thorac Surg 2014;-:-–-
ROSEN ET AL LUNG CANCER SURGERY IN THE NCDB
Table 1. Bivariate Analysis of 30-Day Mortality After Surgical Resection for Non-Small Cell Lung Cancer in 119,146 Patients
Covariate Year of diagnosis 2004 2005 2006 2007 2008 2009 Age group, years < 55 55–64 65–74 75–84 85þ Sex Male Female Race White Nonwhite Insurance status No insurance Private insurance/ managed care Medicaid Medicare Other Income level < $46,000 $46,000þ Unknown Charlson score 0 1 2þ Facility type Community cancer center Comprehensive community cancer center Academic/research program Other, including NCI Facility location Northeast Atlantic Southeast Great Lakes South Midwest West
Number
Dead Within 30 Days (%)
< 0.0001 18,495 19,683 19,878 20,000 20,626 20,464
3.8 3.5 3.5 3.5 3.0 3.0 < 0.0001a
16,319 30,571 43,656 26,403 2,197
1.6 2.1 3.4 5.6 7.1
59,794 59,352
4.4 2.4
105,184 13,962
3.4 3.0
2,310 41,446
3.2 2.2
5,081 67,394 2,915
2.5 4.2 3.6
70,986 41,779 6,381
3.7 2.9 3.0
62,417 41,628 15,101
2.9 3.4 5.0
15,688 61,622
4.2 3.5
40,094 1,742
2.8 3.0
7,146 17,868 27,891 21,722 10,704 9,263 8,535
2.6 2.8 3.5 3.3 4.2 3.4 4.1
Table 1. Continued
Covariate p Value
< 0.0001
0.0061
< 0.0001
< 0.0001
< 0.0001
< 0.0001
< 0.0001
3
Mountain Pacific Facility volume Very low, 1–47 cpy Low, 48–94 cpy Medium, 95–142 cpy High, 143–189 cpy Very high, 190þ cpy Histology Large cell Squamous cell carcinoma Adenocarcinoma BAC Other Primary site Main bronchus Upper lobe Middle lobe Lower lobe Overlapping lesions Lung, not otherwise specified Laterality Right Left Otherb Pathologic stage 1 2 3 4 Unknown Surgical procedure Wedge resection Segmentectomy Lobectomy/bilobectomy Extended lobectomy/ bilobectomyc Pneumonectomy Neoadjuvant treatment None Chemotherapy Radiation Chemoradiation Unknown
Number
Dead Within 30 Days (%)
4,297 11,720
3.1 3.4
94,479 19,636 3,345 607 1,079
3.6 2.8 2.2 1.8 0.7
5,026 36,080 55,461 13,812 8,767
3.6 4.6 2.8 2.2 3.5
1,041 71,378 5,564 36,102 2,209 2,852
5.7 3.2 2.7 3.3 6.4 6.8
68,616 49,108 1,422
3.5 3.0 6.5
66,283 17,434 15,610 4,196 15,623
2.7 3.7 4.0 5.6 4.7
16,293 3,053 85,064 6,181
4.2 3.2 2.6 4.0
8,555
8.5
104,601 2,969 271 4,199 7,106
3.5 3.1 4.8 3.5 0.8
p Value
< 0.0001a
< 0.0001
< 0.0001
< 0.0001
< 0.0001
< 0.0001
0.3588
a
Statistical significance evaluated as a continuous variable by one-way b Other includes main bronchus tumors and analysis of variance. c those with unknown laterality. Includes procedures with chest wall, diaphragmatic, or pericardial resection, or procedures specified as extended, not otherwise specified.
(Continued)
BAC ¼ bronchioloalveolar carcinoma; National Cancer Institute.
cpy ¼ cases per year;
NCI ¼
4
ROSEN ET AL LUNG CANCER SURGERY IN THE NCDB
mortality rates are listed in Table 1. Mortality varied by procedure: lobectomy/bilobectomy, 2.6%; wedge resection, 4.2%; extended lobectomy/bilobectomy, 4.0%; and pneumonectomy, 8.5%. Laterality did impact pneumonectomy mortality, with right-sided mortality (11.1%) being higher than left-sided (6.9%, p < 0.0001). Interestingly, the higher mortality with right-sided pneumonectomy was seen both among patients treated with neoadjuvant therapy (right, 10.6%, and left, 5.9%; p ¼ 0.0031) and patients who were not given neoadjuvant therapy (right, 12.1%, and left, 7.7%; p < 0.0001). Patients undergoing wedge resection were more likely to have a
Ann Thorac Surg 2014;-:-–-
Charlson-Deyo score of 2þ (16.7%) compared with patients undergoing lobectomy/bilobectomy (12.2%) or pneumonectomy (10.0%, p < 0.0001). A multivariable analysis was performed to evaluate the adjusted impact of study variables on mortality (Fig 1). Increasing age, male sex, increasing Charlson-Deyo score, and decreasing facility volume (OR 0.95, 95% CI: 0.93 to 0.97, p < 0.0001, considered as a continuous variable with incremental differences of 10 cases per year) were associated with a higher 30-day operative mortality. Neoadjuvant treatment was not a significant predictor of surgical mortality. In a separate model including only
Fig 1. Results of multivariable logistic regression modeling of 30-day mortality for 112,040 patients who underwent surgical resection for nonsmall cell lung cancer. *Other includes main bronchus tumors and those with unknown laterality. **Extended lobectomy (Ex Lobe/bilobectomy) includes procedures with chest wall, diaphragmatic, or pericardial resection, or procedures specified as extended, not otherwise specified. (CD ¼ Charlson-Deyo; CI ¼ confidence interval; cpy ¼ cases per year; NOS ¼ not otherwise specified.)
Ann Thorac Surg 2014;-:-–-
pneumonectomy patients, right-sided procedures were associated with a greater mortality than left-sided ones (OR 1.77, 95% CI: 1.49 to 2.10, p < 0.0001); however, the use of neoadjuvant chemoradiation therapy (OR 1.04, 95% CI: 0.77 to 1.39, p ¼ 0.81) or neoadjuvant chemotherapy (OR 0.80, 95% CI: 0.55 to 1.17, p ¼ 0.24) did not affect operative mortality.
Extended Length of Stay After Lobectomy The median length of stay was 6 days (IQR, 4 to 9), and varied across procedure types including wedge and segmental resections (median 5 days), lobectomy/bilobectomy (6 days), pneumonectomy (6 days), and extended lobectomy/bilobectomy (7 days). An eLOS of 14 days has been established as a surrogate for significant complications after lobectomy for lung cancer [12]. Overall, 9.1% (or 7,079 patients) had an eLOS after lobectomy/bilobectomy. The incidence of eLOS decreased significantly (p < 0.0001) over the study period, from a high of 9.8% in 2004 to a low of 7.8% in 2009. A multivariable analysis was performed to evaluate predictors of eLOS in patients undergoing lobectomy/ bilobectomy (Fig 2). Male sex, increasing age, increasing Charlson-Deyo score, and right-sided tumors were significant predictors of eLOS. Facility type did not have a significant impact on probability of eLOS, whereas increasing facility volume had a small but significant impact (OR 0.96, 95% CI: 0.94 to 0.98, p ¼ 0.015, considered as a continuous variable with incremental differences of 10 cases per year). Neoadjuvant treatment led to a mixed picture, with neoadjuvant radiation alone increasing risk of eLOS (OR 1.77, 95% CI: 1.12 to 2.80, p ¼ 0.014), and neoadjuvant chemotherapy alone being associated with a decreased risk of eLOS (OR 0.70, 95% CI: 0.57 to 0.86, p ¼ 0.0006). Neoadjuvant chemoradiation was not significantly associated with eLOS rates (OR 0.91, 05% CI: 0.78 to 1.07, p ¼ 0.27).
Impact of Minimally Invasive Approach Recognizing the potential importance of surgical approach (minimally invasive versus thoracotomy), the NCDB added these variables in 2010. We performed a secondary survey of the 2 most recent years in the NCDB to query for approach-specific differences in outcomes. Of the 27,561 NSCLC patients who underwent lobectomy/bilobectomy in 2010 and 2011, 71% were coded as “open/unspecified,” 17% underwent thoracoscopic procedures, 4% underwent robotic procedures, and the remaining 8% were unknown. Table 2 demonstrates differences in 30-day mortality, median length of stay, and incidence of eLOS by operative approach in this cohort. Mortality differed significantly between minimally invasive and open approaches (p < 0.0001); however, there was no significant difference (p ¼ 0.85) between robotic and thoracoscopic procedures. Interestingly, robotic procedures were associated with a higher incidence of eLOS than thoracoscopic procedures (7.1% versus 5.0%, p ¼ 0.0071).
ROSEN ET AL LUNG CANCER SURGERY IN THE NCDB
5
Comment With an annual accrual of more than 119,000 cases of primary lung cancer, the NCDB represents one of the largest data resources for the evaluation of primary lung cancer in the United States (representing a several-fold increase over SEER) [9, 14]. Furthermore, the NCDB does not have the same restrictions as other large administrative databases in terms of geography (SEER), age (Medicare-linked databases), and specialty training of the attending surgeon (STS) [9, 11]. Overall the NCDB cohort appears to be similar to wellestablished administrative data resources (eg, SEERMedicare, STS) for the study of surgically managed NSCLC, but includes a few potentially important differences. Notably, the NCDB contains a greater proportion of female NSCLC patients than SEER-Medicare (49.8% versus 41%), a greater proportion of patients with a Charlson score of 2þ than SEER-Medicare (12.7% versus 7%), and fewer sublobar (ie, wedge and segmental resections) than SEER-Medicare or the STS database (SEER-Medicare 27%, STS 22.5%, NCDB 16.2%) [8, 11].
Mortality The operative mortality rate of 3.4% in the NCDB is similar to the National Inpatient Sample dataset (3.8%) and SEER-Medicare (4%), but is higher than STS (2.3%) [8, 11, 15]. Mortality was higher for wedge resections (4.2%) than for lobectomy/bilobectomies (2.6%), potentially reflecting the poorer health of this patient population, as indicated by their significantly higher Charlson-Deyo scores (16.7% of wedge patients had a score of 2þ compared with 12.2% of lobectomy/bilobectomy patients; p < 0.0001). It is likely that this difference remained significant on multivariable analysis because the NCDB does not capture data such as forced expiratory volume in 1 second or performance status that may further characterize this group of patients as a sicker population compared with patients undergoing lobectomy. The current study identified a relationship between increasing facility volume and lower operative mortality, seen most clearly at the highest volume (190þ cases per year) and lowest volume (1 to 47 cases per year) ends of the facility volume spectrum (0.7% versus 3.6%; p < 0.0001). The volumetric association is similar to data previously reported using the SEER database (3% mortality in the highest volume quintile compared with 6% in the lowest volume quintile) [8]. When facility volume was considered as a continuous variable in an adjusted analysis, the association remained between increasing volume and decreased mortality, but appeared less impressive. That could potentially reflect a threshold effect to volume, with clinically significant skill and safety improvements being achieved at larger volume increments. These results are in line with those of a recent paper that reports volume effects are more exaggerated when examined as quintiles rather than as a continuous variable [7]. Tumor laterality was found to be an important risk factor for perioperative mortality in patients undergoing pneumonectomy in both patients who received
6
ROSEN ET AL LUNG CANCER SURGERY IN THE NCDB
Ann Thorac Surg 2014;-:-–-
Fig 2. Results of multivariable logistic regression modeling of extended length of stay (eLOS) for 73,502 patients who underwent lobectomy/ bilobectomy for non-small cell lung cancer. *Other includes main bronchus tumors and those with unknown laterality. (CD ¼ Charlson-Deyo; CI ¼ confidence interval; cpy ¼ cases per year; NOS ¼ not otherwise specified.)
neoadjuvant treatment and patients who did not. Therefore, it appears that this effect is independent of neoadjuvant treatment status. These results concur with those of several previous studies [16–18], but contrast with others that did not find an association with laterality [19]. The increased mortality risk in right-sided patients has been postulated to reflect (1) a greater increase in right ventricular afterload, theoretically increasing the risk for acute respiratory distress syndrome; and (2) an increased risk of bronchopleural fistula related to a more
tenuous bronchial blood supply (ie, one artery versus two) on the right side [17, 18].
Extended Length of Stay A length of stay longer than 14 days (eLOS) after lobectomy/bilobectomy for NSCLC has previously been shown to correlate with a greater number of adverse events in the postoperative period (3.2 per patient, versus 0.4), and is considered to be a surrogate for significant complications [12]. In the current study, 9.1% of patients had an
Ann Thorac Surg 2014;-:-–-
7
ROSEN ET AL LUNG CANCER SURGERY IN THE NCDB
Table 2. Thirty-Day Mortality Rates and Length of Stay Data by Operative Approach in 25,443 Lobectomy/Bilobectomy Patients With Known Approach Status From 2010 to 2011 Outcome Thirty-day mortality, % Median LOS (IQR) eLOS > 14 days, % a
Open versus thoracoscopic.
eLOS ¼ extended length of stay;
b
Open or Unspecified (n ¼ 19,611)
Thoracoscopic (n ¼ 4,730)
Robotic (n ¼ 1,102)
p Valuea
p Valueb
2.4 6 (4–9) 8.3
1.4 5 (3–7) 5.0
1.4 5 (3–8) 7.1
< 0.0001 . < 0.0001
0.85 . 0.0071
Robotic versus thoracoscopic. IQR ¼ interquartile range;
LOS ¼ length of stay.
eLOS after lobectomy/bilobectomy, which is a bit higher than what has been reported from the STS database (7%) [12]. However, the predictors of eLOS in the NCDB were similar to what has been reported in the STS database (male sex, age, and comorbidities). Increasing hospital case volume correlated with fewer instances of eLOS, which is consistent with a report from the Medicare claims database [20]. Interestingly, the administration of both neoadjuvant chemotherapy and radiotherapy was not a significant predictor of increased eLOS in the NCDB (neoadjuvant chemotherapy was actually associated with a decreased risk of eLOS). These results conflict with a previous studies from the STS that found induction therapy treatment to be associated with an increased risk of eLOS and induction chemotherapy with radiation therapy to be associated with increased major morbidity [12, 21]. The association between neoadjuvant therapy and surgical outcome is likely heavily influenced by the overall health of the patient, and the broader array of patient risk factors within the STS database (eg, preoperative pulmonary function, American Society of Anesthesiologists class, and so forth) may be required to observe this relationship. Furthermore, it is possible that medically fragile patients were more likely to be recognized as poor surgical candidates during their neoadjuvant treatment, thereby removing the weakest patients from this analysis group.
and associates [25]. The relationship between surgical approach and short-term outcome is extremely complex and is worthy of more detailed analysis as the NCDB accumulates additional cases that include the approach variable.
Study Limitations The NCDB does not capture several key data points known to influence complication rates, including detailed comorbidity data and detailed performance status metrics (eg, Zubrod score, American Society of Anesthesiologists score, pulmonary function tests). Furthermore, although the NCDB captures nearly 70% of cancers diagnosed in the United States, it is not a population-based database, and, therefore, the results of this study may not be generalizable to the entire US population. In conclusion, the overall safety profile of surgically managed lung cancer in the NCDB compares well with data from smaller, more restricted administrative databases. The NCDB appears to be a promising data resource to study lung cancer and may prove invaluable to study smaller patient subsets that traditional databases have lacked the power to properly evaluate. This work was supported by the Yale University School of Medicine Medical Student Research Fellowship from the Yale University School of Medicine to JER.
Operative Approach The minimally invasive approach has become increasingly recognized as a safer alternative to traditional lung cancer surgery through thoracotomy, and the NCDB recently added data on the minimally invasive approach to its shared data file beginning in 2010 [22, 23]. To evaluate the importance of approach on perioperative outcome in the NCDB, we performed a subsequent analysis limited to 2010 and 2011 cases undergoing lobectomy/bilobectomy. The proportion of patients undergoing thoracoscopic lobectomies/bilobectomies in the current study (17%) was less than that found in the STS database (more than 40% during similar time period), possibly indicating that specialty-trained thoracic surgeons are more likely to perform procedures using minimally invasive techniques [11, 24]. Overall, minimally invasive approaches appeared to be associated with short-term outcomes superior to those of thoracotomy; however, thoracoscopic and robotic resections appeared to be similar, mirroring a recent report from Kent
References 1. Cruz Dela CS, Tanoue LT, Matthay RA. Lung cancer: epidemiology, etiology, and prevention. Clin Chest Med 2011;32:605–44. 2. Kuzmik GA, Detterbeck FC, Decker RH, et al. Pulmonary resections following prior definitive chemoradiation therapy are associated with acceptable survival. Eur J Cardiothorac Surg 2013;44:66–70. 3. Robinson LA, Ruckdeschel JC, Wagner H, et al. Treatment of non-small cell lung cancer-stage IIIA: ACCP evidence-based clinical practice guidelines (2nd edition). Chest 2007; 132(Suppl):243–65. 4. Posther KE, Harpole DH. The surgical management of lung cancer. Cancer Invest 2006;24:56–67. 5. Howington JA, Blum MG, Chang AC, et al. Treatment of stage I and II non-small cell lung cancer: diagnosis and management of lung cancer, 3rd ed: American College of Chest Physicians evidence-based clinical practice guidelines. Chest 2013;143(Suppl):e278–313. 6. Ramnath N, Dilling TJ, Harris LJ, et al. Treatment of stage III non-small cell lung cancer: diagnosis and management of lung cancer, 3rd ed: American College of Chest Physicians
8
7. 8. 9.
10. 11.
12.
13.
14.
15.
ROSEN ET AL LUNG CANCER SURGERY IN THE NCDB
evidence-based clinical practice guidelines. Chest 2013;143(Suppl):e314–40. Kozower BD, Stukenborg GJ. The relationship between hospital lung cancer resection volume and patient mortality risk. Ann Surg 2011;254:1032–7. Bach PB, Cramer LD, Schrag D, et al. The influence of hospital volume on survival after resection for lung cancer. N Engl J Med 2001;345:181–8. Lerro CC, Robbins AS, Phillips JL, et al. Comparison of cases captured in the national cancer data base with those in population-based central cancer registries. Ann Surg Oncol 2013;20:1759–65. Number of persons by race and Hispanic ethnicity for SEER participants. National Cancer Institute, 2014. Available at: http:// seer.cancer.gov/registries/data.html. Accessed March 17, 2014. Boffa DJ, Allen MS, Grab JD, et al. Data from The Society of Thoracic Surgeons general thoracic surgery database: the surgical management of primary lung tumors. J Thorac Cardiovasc Surg 2008;135:247–54. Wright CD, Gaissert HA, Grab JD, et al. Predictors of prolonged length of stay after lobectomy for lung cancer: a Society of Thoracic Surgeons general thoracic surgery database risk-adjustment model. Ann Thorac Surg 2008;85: 1857–65. Kaboli PJ, Go JT, Hockenberry J, et al. Associations between reduced hospital length of stay and 30-day readmission rate and mortality: 14-year experience in 129 Veterans Affairs hospitals. Ann Intern Med 2012;157:837–45. SEER-Medicare. Number of cases for selected cancers appearing in the data. National Cancer Insitute, 2013. Available at: http://appliedresearch.cancer.gov/seermedicare/ aboutdata/cases.html. Accessed August 29, 2013. Meguid RA, Brooke BS, Chang DC, et al. Are surgical outcomes for lung cancer resections improved at teaching hospitals? Ann Thorac Surg 2008;85:1015–25.
Ann Thorac Surg 2014;-:-–-
16. Martin J, Ginsberg RJ, Abolhoda A, et al. Morbidity and mortality after neoadjuvant therapy for lung cancer: the risks of right pneumonectomy. Ann Thorac Surg 2001;72: 1149–54. 17. Kim AW, Faber LP, Warren WH, et al. Pneumonectomy after chemoradiation therapy for non-small cell lung cancer: does “side” really matter? Ann Thorac Surg 2009;88: 937–44. 18. Fernandez FG, Force SD, Pickens A, et al. Impact of laterality on early and late survival after pneumonectomy. Ann Thorac Surg 2011;92:244–9. 19. Shapiro M, Swanson SJ, Wright CD, et al. Predictors of major morbidity and mortality after pneumonectomy utilizing The Society for Thoracic Surgeons general thoracic surgery database. Ann Thorac Surg 2010;90:927–35. 20. Goodney PP, Stukel TA, Lucas FL, et al. Hospital volume, length of stay, and readmission rates in high-risk surgery. Ann Surg 2003;238:161–7. 21. Kozower BD, Sheng S, O’Brien SM, et al. STS database risk models: predictors of mortality and major morbidity for lung cancer resection. Ann Thorac Surg 2010;90:875–83. 22. Boffa DJ, Dhamija A, Kosinski AS, et al. Fewer complications result from a video-assisted approach to anatomic resection of clinical stage I lung cancer. J Thorac Cardiovasc Surg 2014. In press. 23. Law TD, Boffa DJ, Detterbeck FC, et al. Lethality of cardiovascular events highlights the variable impact of complication type between thoracoscopic and open pulmonary lobectomies. Ann Thorac Surg 2014;97:993–9. 24. Ceppa DP, Kosinski AS, Berry MF, et al. Thoracoscopic lobectomy has increasing benefit in patients with poor pulmonary function. Ann Surg 2012;256:487–93. 25. Kent M, Wang T, Whyte R, et al. Open, video-assisted thoracic surgery, and robotic lobectomy: review of a national database. Ann Thorac Surg 2013;97:236–44.
Background. Surgical resection represents the standard of care for locoregionally confined non-small cell lung cancer (NSCLC); however, surgical complications may compromise the overall outcome. Adverse events after lung cancer surgery have been studied extensively, yet available databases have significant limitations (with respect to size, provider, patient age, and so forth). The National Cancer Database (NCDB) is the largest cancer registry in the world, capturing 67% of newly diagnosed NSCLC in the United States. We studied surgically managed NSCLC patients in the NCDB to more accurately assess factors that influence perioperative outcomes. Methods. Patients diagnosed with NSCLC from 2004 to 2009 in the NCDB who were managed with surgical resection were included (n [ 119,146). Primary endpoints included death within 30 days of surgery and extended length of stay (more than 14 days). Results. Overall 30-day mortality rate was 3.4% and varied by procedure: lobectomy/bilobectomy (2.6%),
wedge resection (4.2%), extended lobectomy/bilobectomy (4%), and pneumonectomy (8.5%). By multivariable analysis, increasing age, male sex, increasing comorbidities, and decreased facility volume were associated with higher 30-day mortality. Of patients who underwent lobectomy, 9.1% had an extended length of stay. On multivariable analysis, increasing age, male sex, increasing comorbidities, decreasing facility volume, and right-sided tumors were associated with increased incidence of extended length of stay. Conclusions. Adverse events after the surgical treatment of NSCLC in the NCDB occur with a similar frequency and are predicted by similar patient, procedural, and facility variables as have been identified by more restricted data resources. The NCDB appears to be a valuable resource to study NSCLC in the United States.
L
28% of the cancers diagnosed in the United States) [9, 10]. Furthermore, the NCDB population lacks many of the restrictions affecting other administrative and clinical databases (eg, SEER Medicare is limited to patients over the age of 65 years [9] and The Society of Thoracic Surgeons (STS) database is almost entirely limited to cases performed by board-certified thoracic surgeons) [11]. Therefore, the NCDB represents an important perspective on the management of NSCLC in the United States. We studied 30-day mortality and extended length of stay (eLOS), as a surrogate for significant but nonlethal complications, in the NCDB to better understand the factors that influence perioperative outcomes in surgically managed NSCLC in the United States.
ung cancer is the leading cause of cancer-related mortality in the United States, claiming more than 160,000 lives in 2010 alone [1]. Surgical resection plays an important role in the management of locoregionally confined non-small cell lung cancer (NSCLC) [2–6]; however, the benefits of surgery may be offset by adverse events in the postoperative period. As a result, short-term surgical outcomes such as death and nonlethal complications are important considerations in comparative effectiveness studies for the management of locoregionally confined NSCLC. Adverse events after lung cancer surgery have been studied extensively, but with mixed, often conflicting results [7, 8]. Because patients, tumors, and cancer providers vary considerably, it is possible that variability in short-term outcome studies to some degree reflects heterogeneity in the available data resources. The National Cancer Data Base (NCDB) represents the largest cancer database in the United States, capturing an estimated 67% of newly diagnosed lung cancers (by comparison, the Surveillance, Epidemiology and End Results [SEER] database only captures approximately Accepted for publication July 7, 2014. Address correspondence to Dr Boffa, 330 Cedar St, BB205, PO Box 208062, New Haven, CT 06520-8062; e-mail: [email protected].
Ó 2014 by The Society of Thoracic Surgeons Published by Elsevier
(Ann Thorac Surg 2014;-:-–-) Ó 2014 by The Society of Thoracic Surgeons
Patients and Methods Data Source The NCDB is a hospital-based tumor registry jointly run by the American College of Surgeons and American Cancer Society since 1985. Currently, all Commission on Cancer accredited facilities are mandated to report all newly diagnosed cancer cases to the NCDB. These approximately 1,445 facilities comprise around 30% of the hospitals in the United States. It is estimated that 67% of 0003-4975/$36.00 http://dx.doi.org/10.1016/j.athoracsur.2014.07.007
2
ROSEN ET AL LUNG CANCER SURGERY IN THE NCDB
lung and bronchus cancers in the United States are captured by the NCDB [9]. The data used in this study are derived from a deidentified NCDB file. The American College of Surgeons and the Commission on Cancer have not verified and are not responsible for the analytic or statistical methodology employed, or the conclusions drawn from these data by the investigator. This study was approved by the Institutional Review Board of the Yale School of Medicine.
Patient Selection The NCDB lung-cancer beta participant user file was queried for all patients over the age of 19 years diagnosed with NSCLC from 2004 to 2009 (n ¼ 668,264). Eligible patients included those undergoing surgical resection (wedge resection, segmentectomy, lobectomy/bilobectomy, extended lobectomy/bilobectomy, pneumonectomy [n ¼ 183,341]). Patients were excluded if they had a noninvasive cancer (n ¼ 133), history of previous cancers (n ¼ 61,370), underwent noncurative intent therapy (n ¼ 1,754), participated in an experimental protocol (n ¼ 404), received intraoperative radiation (n ¼ 728), received hormone therapy (n ¼ 4,100) or immunotherapy (n ¼ 3,118), had an unknown TNM edition number (n ¼ 210), had a pathologic stage of 0 or occult (n ¼ 396), or had missing 30-day mortality data (n ¼ 138). In all, 119,146 patient records met the selection criteria and were used for further analysis. Patients with missing outcome data for length of stay (LOS) were dropped from the analysis of that outcome; furthermore, patients with missing data for neoadjuvant treatment (n ¼ 7,106) were not included in the multivariable models after analyzing for significant confounding differences (data available on request). Additionally, we performed a subsequent analysis of 27,563 separate lobectomy/bilobectomy patients from the 2010 to 2011 NCDB cohort to determine the impact of operative approach on 30-day mortality and eLOS (approach was only added to the data file in 2010). Cases that included both minimally invasive and open approaches were excluded because it is unclear whether they represent a planned thoracoscopic procedure for diagnosis followed by open lobectomy/bilobectomy, or true conversion due to operative findings or complications.
Data Elements The dependent variables included death within 30 days of the primary surgical procedure, and eLOS after the primary surgical procedure; eLOS was defined as more than 14 days from the date of surgery [12]. Independent variables included year of diagnosis, age, sex, race, insurance type, income level, modified Charlson-Deyo score, facility location, facility type, histology, primary tumor site, tumor laterality, pathological stage group, surgical procedure, and neoadjuvant treatment use. Facility volume data for each patient was calculated as the total volume of procedures performed at the treating facility in the year of the patient’s diagnosis, to account for variations in facility volume over time. Comorbidities were represented as a modified Charlson-Deyo score (only includes 0, 1, and 2). Missing data for each variable were coded as
Ann Thorac Surg 2014;-:-–-
“unknown” for the multivariable models, as has been done in similar studies [8]. The use of neoadjuvant radiotherapy was determined by using a specific data field in the NCDB that specifies the temporal relationship of the radiotherapy treatment to the primary surgical procedure. A similar code for systemic therapy was not put in place until 2006. The use of neoadjuvant chemotherapy was determined by comparing the number of days elapsed from diagnosis to surgery, and the number of days elapsed from diagnosis to the start of chemotherapy.
Statistical Analysis Bivariate analysis was performed using the c2 test for categorical covariates and one-way analysis of variance for continuous covariates. Of the 119,146 patients included, 10,887 (9%) were missing data on length of stay and were not included in models of eLOS. Multivariable risk models were developed for each endpoint (30-day mortality, incidence of eLOS). Model development was a multistep process, beginning with simple logistic regression using covariates chosen based on clinical interest and significance in the bivariate analysis (p value threshold < 0.05). Variables were inspected for colinearity. We then transitioned to using generalized linear mixed models using a binomial distribution and logit link with random effects to account for patient clustering within hospitals to calculate parameter estimates and p values [13]. All data elements, except for age and facility volume, were treated as nominal covariates, whereas age and facility volume were analyzed as continuous covariates. Model data are reported using odds ratio (OR), 95% confidence interval (CI), and p values. Model discrimination was assessed using the c-statistic, which was 0.75 for the 30-day mortality model and 0.68 for the eLOS model. All analysis was performed using SAS, version 9.3 (SAS Institute, Cary, NC). A p value of less than 0.05 was set as the threshold for significance.
Results Patient Population In all, 119,146 cases of surgically managed lung cancer were identified and the demographic, cancer, and operative characteristics are described in Table 1. The median age at diagnosis was 67 years, with an interquartile range of 60 to 74. Patients were generally healthy (87% had one or no comorbidities). Most patients underwent lobectomy/bilobectomy (76%), followed by wedge resection (14%), pneumonectomy (7%), and segmentectomy (3%). The median procedure volume of the treating facility was 11 cases per year, with an interquartile range of 5 to 21 cases per year.
Thirty-Day Mortality Overall, 3.4% of patients died within 30 days of their lung cancer surgery, decreasing over the study period from 3.8% in 2004 to 3.0% in 2009 (p < 0.0001). Unadjusted
Ann Thorac Surg 2014;-:-–-
ROSEN ET AL LUNG CANCER SURGERY IN THE NCDB
Table 1. Bivariate Analysis of 30-Day Mortality After Surgical Resection for Non-Small Cell Lung Cancer in 119,146 Patients
Covariate Year of diagnosis 2004 2005 2006 2007 2008 2009 Age group, years < 55 55–64 65–74 75–84 85þ Sex Male Female Race White Nonwhite Insurance status No insurance Private insurance/ managed care Medicaid Medicare Other Income level < $46,000 $46,000þ Unknown Charlson score 0 1 2þ Facility type Community cancer center Comprehensive community cancer center Academic/research program Other, including NCI Facility location Northeast Atlantic Southeast Great Lakes South Midwest West
Number
Dead Within 30 Days (%)
< 0.0001 18,495 19,683 19,878 20,000 20,626 20,464
3.8 3.5 3.5 3.5 3.0 3.0 < 0.0001a
16,319 30,571 43,656 26,403 2,197
1.6 2.1 3.4 5.6 7.1
59,794 59,352
4.4 2.4
105,184 13,962
3.4 3.0
2,310 41,446
3.2 2.2
5,081 67,394 2,915
2.5 4.2 3.6
70,986 41,779 6,381
3.7 2.9 3.0
62,417 41,628 15,101
2.9 3.4 5.0
15,688 61,622
4.2 3.5
40,094 1,742
2.8 3.0
7,146 17,868 27,891 21,722 10,704 9,263 8,535
2.6 2.8 3.5 3.3 4.2 3.4 4.1
Table 1. Continued
Covariate p Value
< 0.0001
0.0061
< 0.0001
< 0.0001
< 0.0001
< 0.0001
< 0.0001
3
Mountain Pacific Facility volume Very low, 1–47 cpy Low, 48–94 cpy Medium, 95–142 cpy High, 143–189 cpy Very high, 190þ cpy Histology Large cell Squamous cell carcinoma Adenocarcinoma BAC Other Primary site Main bronchus Upper lobe Middle lobe Lower lobe Overlapping lesions Lung, not otherwise specified Laterality Right Left Otherb Pathologic stage 1 2 3 4 Unknown Surgical procedure Wedge resection Segmentectomy Lobectomy/bilobectomy Extended lobectomy/ bilobectomyc Pneumonectomy Neoadjuvant treatment None Chemotherapy Radiation Chemoradiation Unknown
Number
Dead Within 30 Days (%)
4,297 11,720
3.1 3.4
94,479 19,636 3,345 607 1,079
3.6 2.8 2.2 1.8 0.7
5,026 36,080 55,461 13,812 8,767
3.6 4.6 2.8 2.2 3.5
1,041 71,378 5,564 36,102 2,209 2,852
5.7 3.2 2.7 3.3 6.4 6.8
68,616 49,108 1,422
3.5 3.0 6.5
66,283 17,434 15,610 4,196 15,623
2.7 3.7 4.0 5.6 4.7
16,293 3,053 85,064 6,181
4.2 3.2 2.6 4.0
8,555
8.5
104,601 2,969 271 4,199 7,106
3.5 3.1 4.8 3.5 0.8
p Value
< 0.0001a
< 0.0001
< 0.0001
< 0.0001
< 0.0001
< 0.0001
0.3588
a
Statistical significance evaluated as a continuous variable by one-way b Other includes main bronchus tumors and analysis of variance. c those with unknown laterality. Includes procedures with chest wall, diaphragmatic, or pericardial resection, or procedures specified as extended, not otherwise specified.
(Continued)
BAC ¼ bronchioloalveolar carcinoma; National Cancer Institute.
cpy ¼ cases per year;
NCI ¼
4
ROSEN ET AL LUNG CANCER SURGERY IN THE NCDB
mortality rates are listed in Table 1. Mortality varied by procedure: lobectomy/bilobectomy, 2.6%; wedge resection, 4.2%; extended lobectomy/bilobectomy, 4.0%; and pneumonectomy, 8.5%. Laterality did impact pneumonectomy mortality, with right-sided mortality (11.1%) being higher than left-sided (6.9%, p < 0.0001). Interestingly, the higher mortality with right-sided pneumonectomy was seen both among patients treated with neoadjuvant therapy (right, 10.6%, and left, 5.9%; p ¼ 0.0031) and patients who were not given neoadjuvant therapy (right, 12.1%, and left, 7.7%; p < 0.0001). Patients undergoing wedge resection were more likely to have a
Ann Thorac Surg 2014;-:-–-
Charlson-Deyo score of 2þ (16.7%) compared with patients undergoing lobectomy/bilobectomy (12.2%) or pneumonectomy (10.0%, p < 0.0001). A multivariable analysis was performed to evaluate the adjusted impact of study variables on mortality (Fig 1). Increasing age, male sex, increasing Charlson-Deyo score, and decreasing facility volume (OR 0.95, 95% CI: 0.93 to 0.97, p < 0.0001, considered as a continuous variable with incremental differences of 10 cases per year) were associated with a higher 30-day operative mortality. Neoadjuvant treatment was not a significant predictor of surgical mortality. In a separate model including only
Fig 1. Results of multivariable logistic regression modeling of 30-day mortality for 112,040 patients who underwent surgical resection for nonsmall cell lung cancer. *Other includes main bronchus tumors and those with unknown laterality. **Extended lobectomy (Ex Lobe/bilobectomy) includes procedures with chest wall, diaphragmatic, or pericardial resection, or procedures specified as extended, not otherwise specified. (CD ¼ Charlson-Deyo; CI ¼ confidence interval; cpy ¼ cases per year; NOS ¼ not otherwise specified.)
Ann Thorac Surg 2014;-:-–-
pneumonectomy patients, right-sided procedures were associated with a greater mortality than left-sided ones (OR 1.77, 95% CI: 1.49 to 2.10, p < 0.0001); however, the use of neoadjuvant chemoradiation therapy (OR 1.04, 95% CI: 0.77 to 1.39, p ¼ 0.81) or neoadjuvant chemotherapy (OR 0.80, 95% CI: 0.55 to 1.17, p ¼ 0.24) did not affect operative mortality.
Extended Length of Stay After Lobectomy The median length of stay was 6 days (IQR, 4 to 9), and varied across procedure types including wedge and segmental resections (median 5 days), lobectomy/bilobectomy (6 days), pneumonectomy (6 days), and extended lobectomy/bilobectomy (7 days). An eLOS of 14 days has been established as a surrogate for significant complications after lobectomy for lung cancer [12]. Overall, 9.1% (or 7,079 patients) had an eLOS after lobectomy/bilobectomy. The incidence of eLOS decreased significantly (p < 0.0001) over the study period, from a high of 9.8% in 2004 to a low of 7.8% in 2009. A multivariable analysis was performed to evaluate predictors of eLOS in patients undergoing lobectomy/ bilobectomy (Fig 2). Male sex, increasing age, increasing Charlson-Deyo score, and right-sided tumors were significant predictors of eLOS. Facility type did not have a significant impact on probability of eLOS, whereas increasing facility volume had a small but significant impact (OR 0.96, 95% CI: 0.94 to 0.98, p ¼ 0.015, considered as a continuous variable with incremental differences of 10 cases per year). Neoadjuvant treatment led to a mixed picture, with neoadjuvant radiation alone increasing risk of eLOS (OR 1.77, 95% CI: 1.12 to 2.80, p ¼ 0.014), and neoadjuvant chemotherapy alone being associated with a decreased risk of eLOS (OR 0.70, 95% CI: 0.57 to 0.86, p ¼ 0.0006). Neoadjuvant chemoradiation was not significantly associated with eLOS rates (OR 0.91, 05% CI: 0.78 to 1.07, p ¼ 0.27).
Impact of Minimally Invasive Approach Recognizing the potential importance of surgical approach (minimally invasive versus thoracotomy), the NCDB added these variables in 2010. We performed a secondary survey of the 2 most recent years in the NCDB to query for approach-specific differences in outcomes. Of the 27,561 NSCLC patients who underwent lobectomy/bilobectomy in 2010 and 2011, 71% were coded as “open/unspecified,” 17% underwent thoracoscopic procedures, 4% underwent robotic procedures, and the remaining 8% were unknown. Table 2 demonstrates differences in 30-day mortality, median length of stay, and incidence of eLOS by operative approach in this cohort. Mortality differed significantly between minimally invasive and open approaches (p < 0.0001); however, there was no significant difference (p ¼ 0.85) between robotic and thoracoscopic procedures. Interestingly, robotic procedures were associated with a higher incidence of eLOS than thoracoscopic procedures (7.1% versus 5.0%, p ¼ 0.0071).
ROSEN ET AL LUNG CANCER SURGERY IN THE NCDB
5
Comment With an annual accrual of more than 119,000 cases of primary lung cancer, the NCDB represents one of the largest data resources for the evaluation of primary lung cancer in the United States (representing a several-fold increase over SEER) [9, 14]. Furthermore, the NCDB does not have the same restrictions as other large administrative databases in terms of geography (SEER), age (Medicare-linked databases), and specialty training of the attending surgeon (STS) [9, 11]. Overall the NCDB cohort appears to be similar to wellestablished administrative data resources (eg, SEERMedicare, STS) for the study of surgically managed NSCLC, but includes a few potentially important differences. Notably, the NCDB contains a greater proportion of female NSCLC patients than SEER-Medicare (49.8% versus 41%), a greater proportion of patients with a Charlson score of 2þ than SEER-Medicare (12.7% versus 7%), and fewer sublobar (ie, wedge and segmental resections) than SEER-Medicare or the STS database (SEER-Medicare 27%, STS 22.5%, NCDB 16.2%) [8, 11].
Mortality The operative mortality rate of 3.4% in the NCDB is similar to the National Inpatient Sample dataset (3.8%) and SEER-Medicare (4%), but is higher than STS (2.3%) [8, 11, 15]. Mortality was higher for wedge resections (4.2%) than for lobectomy/bilobectomies (2.6%), potentially reflecting the poorer health of this patient population, as indicated by their significantly higher Charlson-Deyo scores (16.7% of wedge patients had a score of 2þ compared with 12.2% of lobectomy/bilobectomy patients; p < 0.0001). It is likely that this difference remained significant on multivariable analysis because the NCDB does not capture data such as forced expiratory volume in 1 second or performance status that may further characterize this group of patients as a sicker population compared with patients undergoing lobectomy. The current study identified a relationship between increasing facility volume and lower operative mortality, seen most clearly at the highest volume (190þ cases per year) and lowest volume (1 to 47 cases per year) ends of the facility volume spectrum (0.7% versus 3.6%; p < 0.0001). The volumetric association is similar to data previously reported using the SEER database (3% mortality in the highest volume quintile compared with 6% in the lowest volume quintile) [8]. When facility volume was considered as a continuous variable in an adjusted analysis, the association remained between increasing volume and decreased mortality, but appeared less impressive. That could potentially reflect a threshold effect to volume, with clinically significant skill and safety improvements being achieved at larger volume increments. These results are in line with those of a recent paper that reports volume effects are more exaggerated when examined as quintiles rather than as a continuous variable [7]. Tumor laterality was found to be an important risk factor for perioperative mortality in patients undergoing pneumonectomy in both patients who received
6
ROSEN ET AL LUNG CANCER SURGERY IN THE NCDB
Ann Thorac Surg 2014;-:-–-
Fig 2. Results of multivariable logistic regression modeling of extended length of stay (eLOS) for 73,502 patients who underwent lobectomy/ bilobectomy for non-small cell lung cancer. *Other includes main bronchus tumors and those with unknown laterality. (CD ¼ Charlson-Deyo; CI ¼ confidence interval; cpy ¼ cases per year; NOS ¼ not otherwise specified.)
neoadjuvant treatment and patients who did not. Therefore, it appears that this effect is independent of neoadjuvant treatment status. These results concur with those of several previous studies [16–18], but contrast with others that did not find an association with laterality [19]. The increased mortality risk in right-sided patients has been postulated to reflect (1) a greater increase in right ventricular afterload, theoretically increasing the risk for acute respiratory distress syndrome; and (2) an increased risk of bronchopleural fistula related to a more
tenuous bronchial blood supply (ie, one artery versus two) on the right side [17, 18].
Extended Length of Stay A length of stay longer than 14 days (eLOS) after lobectomy/bilobectomy for NSCLC has previously been shown to correlate with a greater number of adverse events in the postoperative period (3.2 per patient, versus 0.4), and is considered to be a surrogate for significant complications [12]. In the current study, 9.1% of patients had an
Ann Thorac Surg 2014;-:-–-
7
ROSEN ET AL LUNG CANCER SURGERY IN THE NCDB
Table 2. Thirty-Day Mortality Rates and Length of Stay Data by Operative Approach in 25,443 Lobectomy/Bilobectomy Patients With Known Approach Status From 2010 to 2011 Outcome Thirty-day mortality, % Median LOS (IQR) eLOS > 14 days, % a
Open versus thoracoscopic.
eLOS ¼ extended length of stay;
b
Open or Unspecified (n ¼ 19,611)
Thoracoscopic (n ¼ 4,730)
Robotic (n ¼ 1,102)
p Valuea
p Valueb
2.4 6 (4–9) 8.3
1.4 5 (3–7) 5.0
1.4 5 (3–8) 7.1
< 0.0001 . < 0.0001
0.85 . 0.0071
Robotic versus thoracoscopic. IQR ¼ interquartile range;
LOS ¼ length of stay.
eLOS after lobectomy/bilobectomy, which is a bit higher than what has been reported from the STS database (7%) [12]. However, the predictors of eLOS in the NCDB were similar to what has been reported in the STS database (male sex, age, and comorbidities). Increasing hospital case volume correlated with fewer instances of eLOS, which is consistent with a report from the Medicare claims database [20]. Interestingly, the administration of both neoadjuvant chemotherapy and radiotherapy was not a significant predictor of increased eLOS in the NCDB (neoadjuvant chemotherapy was actually associated with a decreased risk of eLOS). These results conflict with a previous studies from the STS that found induction therapy treatment to be associated with an increased risk of eLOS and induction chemotherapy with radiation therapy to be associated with increased major morbidity [12, 21]. The association between neoadjuvant therapy and surgical outcome is likely heavily influenced by the overall health of the patient, and the broader array of patient risk factors within the STS database (eg, preoperative pulmonary function, American Society of Anesthesiologists class, and so forth) may be required to observe this relationship. Furthermore, it is possible that medically fragile patients were more likely to be recognized as poor surgical candidates during their neoadjuvant treatment, thereby removing the weakest patients from this analysis group.
and associates [25]. The relationship between surgical approach and short-term outcome is extremely complex and is worthy of more detailed analysis as the NCDB accumulates additional cases that include the approach variable.
Study Limitations The NCDB does not capture several key data points known to influence complication rates, including detailed comorbidity data and detailed performance status metrics (eg, Zubrod score, American Society of Anesthesiologists score, pulmonary function tests). Furthermore, although the NCDB captures nearly 70% of cancers diagnosed in the United States, it is not a population-based database, and, therefore, the results of this study may not be generalizable to the entire US population. In conclusion, the overall safety profile of surgically managed lung cancer in the NCDB compares well with data from smaller, more restricted administrative databases. The NCDB appears to be a promising data resource to study lung cancer and may prove invaluable to study smaller patient subsets that traditional databases have lacked the power to properly evaluate. This work was supported by the Yale University School of Medicine Medical Student Research Fellowship from the Yale University School of Medicine to JER.
Operative Approach The minimally invasive approach has become increasingly recognized as a safer alternative to traditional lung cancer surgery through thoracotomy, and the NCDB recently added data on the minimally invasive approach to its shared data file beginning in 2010 [22, 23]. To evaluate the importance of approach on perioperative outcome in the NCDB, we performed a subsequent analysis limited to 2010 and 2011 cases undergoing lobectomy/bilobectomy. The proportion of patients undergoing thoracoscopic lobectomies/bilobectomies in the current study (17%) was less than that found in the STS database (more than 40% during similar time period), possibly indicating that specialty-trained thoracic surgeons are more likely to perform procedures using minimally invasive techniques [11, 24]. Overall, minimally invasive approaches appeared to be associated with short-term outcomes superior to those of thoracotomy; however, thoracoscopic and robotic resections appeared to be similar, mirroring a recent report from Kent
References 1. Cruz Dela CS, Tanoue LT, Matthay RA. Lung cancer: epidemiology, etiology, and prevention. Clin Chest Med 2011;32:605–44. 2. Kuzmik GA, Detterbeck FC, Decker RH, et al. Pulmonary resections following prior definitive chemoradiation therapy are associated with acceptable survival. Eur J Cardiothorac Surg 2013;44:66–70. 3. Robinson LA, Ruckdeschel JC, Wagner H, et al. Treatment of non-small cell lung cancer-stage IIIA: ACCP evidence-based clinical practice guidelines (2nd edition). Chest 2007; 132(Suppl):243–65. 4. Posther KE, Harpole DH. The surgical management of lung cancer. Cancer Invest 2006;24:56–67. 5. Howington JA, Blum MG, Chang AC, et al. Treatment of stage I and II non-small cell lung cancer: diagnosis and management of lung cancer, 3rd ed: American College of Chest Physicians evidence-based clinical practice guidelines. Chest 2013;143(Suppl):e278–313. 6. Ramnath N, Dilling TJ, Harris LJ, et al. Treatment of stage III non-small cell lung cancer: diagnosis and management of lung cancer, 3rd ed: American College of Chest Physicians
8
7. 8. 9.
10. 11.
12.
13.
14.
15.
ROSEN ET AL LUNG CANCER SURGERY IN THE NCDB
evidence-based clinical practice guidelines. Chest 2013;143(Suppl):e314–40. Kozower BD, Stukenborg GJ. The relationship between hospital lung cancer resection volume and patient mortality risk. Ann Surg 2011;254:1032–7. Bach PB, Cramer LD, Schrag D, et al. The influence of hospital volume on survival after resection for lung cancer. N Engl J Med 2001;345:181–8. Lerro CC, Robbins AS, Phillips JL, et al. Comparison of cases captured in the national cancer data base with those in population-based central cancer registries. Ann Surg Oncol 2013;20:1759–65. Number of persons by race and Hispanic ethnicity for SEER participants. National Cancer Institute, 2014. Available at: http:// seer.cancer.gov/registries/data.html. Accessed March 17, 2014. Boffa DJ, Allen MS, Grab JD, et al. Data from The Society of Thoracic Surgeons general thoracic surgery database: the surgical management of primary lung tumors. J Thorac Cardiovasc Surg 2008;135:247–54. Wright CD, Gaissert HA, Grab JD, et al. Predictors of prolonged length of stay after lobectomy for lung cancer: a Society of Thoracic Surgeons general thoracic surgery database risk-adjustment model. Ann Thorac Surg 2008;85: 1857–65. Kaboli PJ, Go JT, Hockenberry J, et al. Associations between reduced hospital length of stay and 30-day readmission rate and mortality: 14-year experience in 129 Veterans Affairs hospitals. Ann Intern Med 2012;157:837–45. SEER-Medicare. Number of cases for selected cancers appearing in the data. National Cancer Insitute, 2013. Available at: http://appliedresearch.cancer.gov/seermedicare/ aboutdata/cases.html. Accessed August 29, 2013. Meguid RA, Brooke BS, Chang DC, et al. Are surgical outcomes for lung cancer resections improved at teaching hospitals? Ann Thorac Surg 2008;85:1015–25.
Ann Thorac Surg 2014;-:-–-
16. Martin J, Ginsberg RJ, Abolhoda A, et al. Morbidity and mortality after neoadjuvant therapy for lung cancer: the risks of right pneumonectomy. Ann Thorac Surg 2001;72: 1149–54. 17. Kim AW, Faber LP, Warren WH, et al. Pneumonectomy after chemoradiation therapy for non-small cell lung cancer: does “side” really matter? Ann Thorac Surg 2009;88: 937–44. 18. Fernandez FG, Force SD, Pickens A, et al. Impact of laterality on early and late survival after pneumonectomy. Ann Thorac Surg 2011;92:244–9. 19. Shapiro M, Swanson SJ, Wright CD, et al. Predictors of major morbidity and mortality after pneumonectomy utilizing The Society for Thoracic Surgeons general thoracic surgery database. Ann Thorac Surg 2010;90:927–35. 20. Goodney PP, Stukel TA, Lucas FL, et al. Hospital volume, length of stay, and readmission rates in high-risk surgery. Ann Surg 2003;238:161–7. 21. Kozower BD, Sheng S, O’Brien SM, et al. STS database risk models: predictors of mortality and major morbidity for lung cancer resection. Ann Thorac Surg 2010;90:875–83. 22. Boffa DJ, Dhamija A, Kosinski AS, et al. Fewer complications result from a video-assisted approach to anatomic resection of clinical stage I lung cancer. J Thorac Cardiovasc Surg 2014. In press. 23. Law TD, Boffa DJ, Detterbeck FC, et al. Lethality of cardiovascular events highlights the variable impact of complication type between thoracoscopic and open pulmonary lobectomies. Ann Thorac Surg 2014;97:993–9. 24. Ceppa DP, Kosinski AS, Berry MF, et al. Thoracoscopic lobectomy has increasing benefit in patients with poor pulmonary function. Ann Surg 2012;256:487–93. 25. Kent M, Wang T, Whyte R, et al. Open, video-assisted thoracic surgery, and robotic lobectomy: review of a national database. Ann Thorac Surg 2013;97:236–44.